Two-Sample Treatment Effect Estimands under Nonproportional Hazards with Right-Censored Data

戴以誠 博士

Dr. Yi-Cheng Tai

Department of Statistics and Data Science, Cornell University

Abstract

With survival outcome data, the hazard ratio is traditionally considered as the gold standard to quantify the treatment effects. However, the advancements in immunotherapy for clinical oncology challenge the proportional hazards assumption, making the interpretation of hazard ratios questionable. While restricted mean survival time has emerged as an alternative, and other model-based estimation and testing methods have been proposed, there is still no consensus on the best approach to summarize the treatment effect under nonproportional hazards. In this talk, I will discuss the issues associated with nonproportional hazards in treatment effect quantification and introduce a novel two-sample treatment effect estimand to assess the relative performance of two groups over time. Our method includes a graphical tool to trace treatment progression and align relative changes in hazard functions, depicting the instantaneous effects of treatments. To enhance efficiency and incorporate baseline covariates, we adopt a double machine learning framework, leveraging modern machine learning techniques to ensure robust and valid inference.

日  期:114年1月8日(星期三) 10:00~11:00

地  點：本校數學館312教室（嘉義縣民雄鄉大學路168號）

歡迎參加   敬請公佈

交通資訊及校內地圖